[postgis-devel] Dataset metadata

[Darafei "Komяpa" Praliaskouski]

Hi!

It looks there are a bunch of settings people want to set from time to time
on per-dataset level, and have access to those in processing routines.

What has come to my mind:
 - SRID, to store it once for everything and make geometry POINT even more
lightweight;
 - Dimensionality choices, so that ST_Intersection() of two adjacent
polygons can give POLYGON EMPTY instead of shared LINESTRING;
 - Precision, to be able to distinguish slivers in geometry from meaningful
parts;
 - Precision, to segmentize curves to produce results no worse than;
 - Precision, to automatically apply ST_QuantizeCoordinates,
ST_RemoveRepeatedPoints and ST_Simplify;
 - Invalid geometries repair options: do we drop holes outside shell or
make them outers, do we fail on ST_Buffer(butterfly, 0) or let it
re-polygonize the input, do we need the lines that holes share or they can
be dissolved;
 - Store the "ESRI flag" from
https://postgis.net/docs/manual-dev/ST_IsValidDetail.html;
 - Meaning of M, if any - is it weight, is it line reference length, is it
time and objects are trajectories?

Are there any other metadata you've seen the need to store per dataset?

What are our storage options?

I imagine it can be done as some cross of spatial_ref_sys +
geometry_columns, where we can store an ID of some set of options in
geometry itself, but that still requires an ID in the tuple.

Typmod can be used, but - is it preserved reliably enough?

Any other thoughts? :)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20180316/6f0977f5/attachment.html>